Assignments Policies You must typeset, choices: Word (very easy to type math expressions) Latex (very easy to type math expressions) Google doc Plain text + math formula Your favorite text/doc editor Submit a single PDF for the written part The could be some coding tasks in assignments Unless otherwise stated, you don t need to submit your code for HW assignments unless explicitly told to do so Follow instructions given in each assignment Collaboration is encouraged among small groups (e.g., 2-3 students). You should disclose with whom you have discussed you problems This helps especially if multiple people make identical mistakes
CS 460/560 Introduction to Computational Robotics Fall 2017, Rutgers University Lecture 04 Sensor Mechanisms Instructor: Jingjin Yu
Outline Common sensors Encoders Heading/orientation Acceleration Inertial measurement unit (IMU) Range sensors Motion/speed sensors Vision Modeling sensors
Encoders, Absolute Wheel encoders measure position and rotation speed Absolute (binary, gray code) 010 001 011 000 100 111 101 110 Measure exact (quantized) wheel position Do not need to track time for position With time, can also measure speed t 13 track Gray code absolute rotary encoder Images from wikipedia
Encoders, Relative Relative Index t Relative: mainly direction/speed A single set of gaps may have difficulty determining direction Can add an offset set of gaps to address this Offset by ¼, forward and backward become asymmetric Images from IMTR, www.phidgets.com/docs21/encoder_primer
Encoders, Applications Encoders are very useful in robotics For figuring out wheel states For figuring out positions of robotic arms E.g., Universal Robots UR5 arm Can be highly accurate! For UR5, the repeatability is ±0.1 mm or ±0.0039 in
Orientation Sensors Generally, there are two types of orientation sensors 2D: SE 2 = R 2 S 1, single orientation on S 1 3D: SE 3 = R 3 SO(3), three orientations, e.g., (yaw, pitch, roll) Two dimensional orientation sensor Compasses! They don t always work reliably Globally, magnetic field changes over time Locally Natural magnetic objects Man made magnetic/electro-magnetic fields Magnetic field direction image from geomag.org/info/declination.html
Orientation Sensors, Continued Gyroscopes (three dimensional orientation sensors) Operational principle: angular momentum (inertia) A spinning wheel likes to stay the way it is A gimbal is a device allowing free rotation along one axis A gyro is a wheel mounted on 2+ gimbals Used in many equipment, e.g., aircraft Can be non-mechanical https://www.youtube.com/watch?v=cquva_ipesa A toy gyro Gimbal camera mount 3-gimbal gyro Gyroscope image from wikipedia
Acceleration Sensor Accelerometers Also based on inertia (linear momentum) Newton s first law A mass tends to keep its current momentum Changing it requires force F = ma = kx (Newton s second law) Measures a single direction, e.g. x For three dimensions, need three such mechanisms Can be done with a single mass though A useful trick for figuring out where the ground is Earth has gravity, which is an acceleration force Actually very large and fixed magnitude, often dominates other forces Provides us with a ground direction Can be made very small! May replace spring with piezoelectric mechanism
Inertial Measurement Unit (IMU) IMU uses both angular and linear inertial effects How? Gyro + accelerometer Produces position, velocity, acceleration Pipeline IMUs may also have magnetic and other sensors
Range Sensors Time-of-flight ranging Measure distance by measuring time traveled (by sound or light) distance = speed travel time Light travels much faster Accurate, but more difficult to measure Sound is slower, but works in water well Many things to consider Time measurement Speed variance (esp. sound) Signal uniformity (e.g., sound cone) Triangulation (will cover this later) Phase shift Measure distance by measuring phase of light Light is a wave Sonar Lidar (often uses triangulation)
Range Sensors, Continued Sonars Laser scanners Two dimensional Three dimensional Similar principles Other (more complex) methods Structured light Used in Microsoft Kinect iphone X has something similar See IAMR 4.1
TED version: https://www.youtube.com/watch?v=y_9vd4hwlva Speed Sensors and Vision Sensors Speed sensor: Doppler effect Uses wave aspect of sound/light Faster then more frequent arrival Very accurate Vison sensors You have seen many of these Yes just cameras But can do 1T FPS now
Sensor Modeling A sensor can be modeled using a function h: X: state space Y: observation space h: sensor (mapping) function h: X Y Often, X and Y may be the same; but not always Examples Position sensor: h: R 3 R 3 Orientation sensor: SO(3) SO(3) Contact sensor: h: SE(2) {0,1} We will not do this very formally in this class